#include "internal.h" static int call_user(int type, struct mg_connection *conn, void *p) { if (conn != NULL && conn->ctx != NULL) { conn->event.user_data = conn->ctx->user_data; conn->event.type = type; conn->event.event_param = p; conn->event.request_info = &conn->request_info; conn->event.conn = conn; } return conn == NULL || conn->ctx == NULL || conn->ctx->event_handler == NULL ? 0 : conn->ctx->event_handler(&conn->event); } static int mg_socketpair(sock_t sp[2]) { struct sockaddr_in sa; sock_t sock, ret = -1; socklen_t len = sizeof(sa); sp[0] = sp[1] = INVALID_SOCKET; (void) memset(&sa, 0, sizeof(sa)); sa.sin_family = AF_INET; sa.sin_port = htons(0); sa.sin_addr.s_addr = htonl(0x7f000001); if ((sock = socket(AF_INET, SOCK_STREAM, 0)) != INVALID_SOCKET && !bind(sock, (struct sockaddr *) &sa, len) && !listen(sock, 1) && !getsockname(sock, (struct sockaddr *) &sa, &len) && (sp[0] = socket(AF_INET, SOCK_STREAM, 6)) != -1 && !connect(sp[0], (struct sockaddr *) &sa, len) && (sp[1] = accept(sock,(struct sockaddr *) &sa, &len)) != INVALID_SOCKET) { set_close_on_exec(sp[0]); set_close_on_exec(sp[1]); ret = 0; } else { if (sp[0] != INVALID_SOCKET) closesocket(sp[0]); if (sp[1] != INVALID_SOCKET) closesocket(sp[1]); sp[0] = sp[1] = INVALID_SOCKET; } closesocket(sock); return ret; } static FILE *mg_fopen(const char *path, const char *mode) { #ifdef _WIN32 wchar_t wbuf[PATH_MAX], wmode[20]; to_unicode(path, wbuf, ARRAY_SIZE(wbuf)); MultiByteToWideChar(CP_UTF8, 0, mode, -1, wmode, ARRAY_SIZE(wmode)); return _wfopen(wbuf, wmode); #else return fopen(path, mode); #endif } // Print error message to the opened error log stream. static void cry(struct mg_connection *conn, const char *fmt, ...) { char buf[MG_BUF_LEN], src_addr[IP_ADDR_STR_LEN]; va_list ap; FILE *fp; time_t timestamp; va_start(ap, fmt); (void) vsnprintf(buf, sizeof(buf), fmt, ap); va_end(ap); // Do not lock when getting the callback value, here and below. // I suppose this is fine, since function cannot disappear in the // same way string option can. if (call_user(MG_EVENT_LOG, conn, buf) == 0) { fp = conn->ctx == NULL || conn->ctx->config[ERROR_LOG_FILE] == NULL ? NULL : fopen(conn->ctx->config[ERROR_LOG_FILE], "a+"); if (fp != NULL) { flockfile(fp); timestamp = time(NULL); sockaddr_to_string(src_addr, sizeof(src_addr), &conn->client.rsa); fprintf(fp, "[%010lu] [error] [client %s] ", (unsigned long) timestamp, src_addr); if (conn->request_info.request_method != NULL) { fprintf(fp, "%s %s: ", conn->request_info.request_method, conn->request_info.uri); } fprintf(fp, "%s", buf); fputc('\n', fp); funlockfile(fp); fclose(fp); } } } const char *mg_version(void) { return MONGOOSE_VERSION; } // HTTP 1.1 assumes keep alive if "Connection:" header is not set // This function must tolerate situations when connection info is not // set up, for example if request parsing failed. static int should_keep_alive(const struct mg_connection *conn) { const char *http_version = conn->request_info.http_version; const char *header = mg_get_header(conn, "Connection"); if (conn->must_close || conn->status_code == 401 || mg_strcasecmp(conn->ctx->config[ENABLE_KEEP_ALIVE], "yes") != 0 || (header != NULL && mg_strcasecmp(header, "keep-alive") != 0) || (header == NULL && http_version && strcmp(http_version, "1.1"))) { return 0; } return 1; } static const char *suggest_connection_header(const struct mg_connection *conn) { return should_keep_alive(conn) ? "keep-alive" : "close"; } static void send_http_error(struct mg_connection *conn, int status, const char *reason, const char *fmt, ...) { char buf[MG_BUF_LEN]; va_list ap; int len = 0; conn->status_code = status; buf[0] = '\0'; // Errors 1xx, 204 and 304 MUST NOT send a body if (status > 199 && status != 204 && status != 304) { len = mg_snprintf(buf, sizeof(buf), "Error %d: %s", status, reason); buf[len++] = '\n'; va_start(ap, fmt); len += mg_vsnprintf(buf + len, sizeof(buf) - len, fmt, ap); va_end(ap); } DEBUG_TRACE(("[%s]", buf)); if (call_user(MG_HTTP_ERROR, conn, (void *) (long) status) == 0) { mg_printf(conn, "HTTP/1.1 %d %s\r\n" "Content-Length: %d\r\n" "Connection: %s\r\n\r\n", status, reason, len, suggest_connection_header(conn)); conn->num_bytes_sent += mg_printf(conn, "%s", buf); } if (status >= 500) { conn->must_close = 1; } } // Return 1 if real file has been found, 0 otherwise static int convert_uri_to_file_name(struct mg_connection *conn, char *buf, size_t buf_len, struct file *filep) { struct vec a, b; const char *rewrite, *uri = conn->request_info.uri, *root = conn->ctx->config[DOCUMENT_ROOT]; char *p; int match_len; char gz_path[PATH_MAX]; char const* accept_encoding; // No filesystem access if (root == NULL) { return 0; } // Using buf_len - 1 because memmove() for PATH_INFO may shift part // of the path one byte on the right. // If document_root is NULL, leave the file empty. mg_snprintf(buf, buf_len - 1, "%s%s", root, uri); rewrite = conn->ctx->config[REWRITE]; while ((rewrite = next_option(rewrite, &a, &b)) != NULL) { if ((match_len = match_prefix(a.ptr, a.len, uri)) > 0) { mg_snprintf(buf, buf_len - 1, "%.*s%s", (int) b.len, b.ptr, uri + match_len); break; } } if (mg_stat(buf, filep)) { return 1; } // if we can't find the actual file, look for the file // with the same name but a .gz extension. If we find it, // use that and set the gzipped flag in the file struct // to indicate that the response need to have the content- // encoding: gzip header // we can only do this if the browser declares support if ((accept_encoding = mg_get_header(conn, "Accept-Encoding")) != NULL) { if (strstr(accept_encoding,"gzip") != NULL) { snprintf(gz_path, sizeof(gz_path), "%s.gz", buf); if (mg_stat(gz_path, filep)) { filep->gzipped = 1; return 1; } } } // Support PATH_INFO for CGI scripts. for (p = buf + strlen(root == NULL ? "" : root); *p != '\0'; p++) { if (*p == '/') { *p = '\0'; if (match_prefix(conn->ctx->config[CGI_EXTENSIONS], strlen(conn->ctx->config[CGI_EXTENSIONS]), buf) > 0 && mg_stat(buf, filep)) { // Shift PATH_INFO block one character right, e.g. // "/x.cgi/foo/bar\x00" => "/x.cgi\x00/foo/bar\x00" // conn->path_info is pointing to the local variable "path" declared // in handle_request(), so PATH_INFO is not valid after // handle_request returns. conn->path_info = p + 1; memmove(p + 2, p + 1, strlen(p + 1) + 1); // +1 is for trailing \0 p[1] = '/'; return 1; } else { *p = '/'; } } } return 0; } static void construct_etag(char *buf, size_t buf_len, const struct file *filep) { snprintf(buf, buf_len, "\"%lx.%" INT64_FMT "\"", (unsigned long) filep->modification_time, filep->size); } static void fclose_on_exec(FILE *fp) { if (fp != NULL) { #ifndef _WIN32 fcntl(fileno(fp), F_SETFD, FD_CLOEXEC); #endif } } static void handle_file_request(struct mg_connection *conn, const char *path, struct file *filep) { char date[64], lm[64], etag[64], range[64]; const char *msg = "OK", *hdr; time_t curtime = time(NULL); int64_t cl, r1, r2; struct vec mime_vec; int n; char gz_path[PATH_MAX]; char const* encoding = ""; FILE *fp; get_mime_type(conn->ctx, path, &mime_vec); cl = filep->size; conn->status_code = 200; range[0] = '\0'; // if this file is in fact a pre-gzipped file, rewrite its filename // it's important to rewrite the filename after resolving // the mime type from it, to preserve the actual file's type if (filep->gzipped) { snprintf(gz_path, sizeof(gz_path), "%s.gz", path); path = gz_path; encoding = "Content-Encoding: gzip\r\n"; } if ((fp = mg_fopen(path, "rb")) == NULL) { send_http_error(conn, 500, http_500_error, "fopen(%s): %s", path, strerror(ERRNO)); return; } fclose_on_exec(fp); // If Range: header specified, act accordingly r1 = r2 = 0; hdr = mg_get_header(conn, "Range"); if (hdr != NULL && (n = parse_range_header(hdr, &r1, &r2)) > 0 && r1 >= 0 && r2 >= 0) { // actually, range requests don't play well with a pre-gzipped // file (since the range is specified in the uncmpressed space) if (filep->gzipped) { send_http_error(conn, 501, "Not Implemented", "range requests in gzipped files are not supported"); return; } conn->status_code = 206; cl = n == 2 ? (r2 > cl ? cl : r2) - r1 + 1: cl - r1; mg_snprintf(range, sizeof(range), "Content-Range: bytes " "%" INT64_FMT "-%" INT64_FMT "/%" INT64_FMT "\r\n", r1, r1 + cl - 1, filep->size); msg = "Partial Content"; } // Prepare Etag, Date, Last-Modified headers. Must be in UTC, according to // http://www.w3.org/Protocols/rfc2616/rfc2616-sec3.html#sec3.3 gmt_time_string(date, sizeof(date), &curtime); gmt_time_string(lm, sizeof(lm), &filep->modification_time); construct_etag(etag, sizeof(etag), filep); (void) mg_printf(conn, "HTTP/1.1 %d %s\r\n" "Date: %s\r\n" "Last-Modified: %s\r\n" "Etag: %s\r\n" "Content-Type: %.*s\r\n" "Content-Length: %" INT64_FMT "\r\n" "Connection: %s\r\n" "Accept-Ranges: bytes\r\n" "%s%s%s\r\n", conn->status_code, msg, date, lm, etag, (int) mime_vec.len, mime_vec.ptr, cl, suggest_connection_header(conn), range, encoding, EXTRA_HTTP_HEADERS); if (strcmp(conn->request_info.request_method, "HEAD") != 0) { send_file_data(conn, fp, r1, cl); } fclose(fp); } void mg_send_file(struct mg_connection *conn, const char *path) { struct file file = STRUCT_FILE_INITIALIZER; if (mg_stat(path, &file)) { handle_file_request(conn, path, &file); } else { send_http_error(conn, 404, "Not Found", "%s", "File not found"); } } // For given directory path, substitute it to valid index file. // Return 0 if index file has been found, -1 if not found. // If the file is found, it's stats is returned in stp. static int substitute_index_file(struct mg_connection *conn, char *path, size_t path_len, struct file *filep) { const char *list = conn->ctx->config[INDEX_FILES]; struct file file = STRUCT_FILE_INITIALIZER; struct vec filename_vec; size_t n = strlen(path); int found = 0; // The 'path' given to us points to the directory. Remove all trailing // directory separator characters from the end of the path, and // then append single directory separator character. while (n > 0 && path[n - 1] == '/') { n--; } path[n] = '/'; // Traverse index files list. For each entry, append it to the given // path and see if the file exists. If it exists, break the loop while ((list = next_option(list, &filename_vec, NULL)) != NULL) { // Ignore too long entries that may overflow path buffer if (filename_vec.len > path_len - (n + 2)) continue; // Prepare full path to the index file mg_strlcpy(path + n + 1, filename_vec.ptr, filename_vec.len + 1); // Does it exist? if (mg_stat(path, &file)) { // Yes it does, break the loop *filep = file; found = 1; break; } } // If no index file exists, restore directory path if (!found) { path[n] = '\0'; } return found; } // Return True if we should reply 304 Not Modified. static int is_not_modified(const struct mg_connection *conn, const struct file *filep) { char etag[64]; const char *ims = mg_get_header(conn, "If-Modified-Since"); const char *inm = mg_get_header(conn, "If-None-Match"); construct_etag(etag, sizeof(etag), filep); return (inm != NULL && !mg_strcasecmp(etag, inm)) || (ims != NULL && filep->modification_time <= parse_date_string(ims)); } static void mkcol(struct mg_connection *conn, const char *path) { int rc, body_len; struct de de; memset(&de.file, 0, sizeof(de.file)); mg_stat(path, &de.file); if (de.file.modification_time) { send_http_error(conn, 405, "Method Not Allowed", "mkcol(%s): %s", path, strerror(ERRNO)); return; } body_len = conn->data_len - conn->request_len; if(body_len > 0) { send_http_error(conn, 415, "Unsupported media type", "mkcol(%s): %s", path, strerror(ERRNO)); return; } rc = mg_mkdir(path, 0755); if (rc == 0) { conn->status_code = 201; mg_printf(conn, "HTTP/1.1 %d Created\r\n\r\n", conn->status_code); } else if (rc == -1) { if(errno == EEXIST) send_http_error(conn, 405, "Method Not Allowed", "mkcol(%s): %s", path, strerror(ERRNO)); else if(errno == EACCES) send_http_error(conn, 403, "Forbidden", "mkcol(%s): %s", path, strerror(ERRNO)); else if(errno == ENOENT) send_http_error(conn, 409, "Conflict", "mkcol(%s): %s", path, strerror(ERRNO)); else send_http_error(conn, 500, http_500_error, "fopen(%s): %s", path, strerror(ERRNO)); } } static void put_file(struct mg_connection *conn, const char *path) { struct file file = STRUCT_FILE_INITIALIZER; FILE *fp; const char *range; int64_t r1, r2; int rc; conn->status_code = mg_stat(path, &file) ? 200 : 201; if ((rc = put_dir(path)) == 0) { mg_printf(conn, "HTTP/1.1 %d OK\r\n\r\n", conn->status_code); } else if (rc == -1) { send_http_error(conn, 500, http_500_error, "put_dir(%s): %s", path, strerror(ERRNO)); } else if ((fp = mg_fopen(path, "wb+")) == NULL) { fclose(fp); send_http_error(conn, 500, http_500_error, "fopen(%s): %s", path, strerror(ERRNO)); } else { fclose_on_exec(fp); range = mg_get_header(conn, "Content-Range"); r1 = r2 = 0; if (range != NULL && parse_range_header(range, &r1, &r2) > 0) { conn->status_code = 206; fseeko(fp, r1, SEEK_SET); } if (!forward_body_data(conn, fp, INVALID_SOCKET, NULL)) { conn->status_code = 500; } mg_printf(conn, "HTTP/1.1 %d OK\r\nContent-Length: 0\r\n\r\n", conn->status_code); fclose(fp); } } static void send_ssi_file(struct mg_connection *, const char *, FILE *, int); static void do_ssi_include(struct mg_connection *conn, const char *ssi, char *tag, int include_level) { char file_name[MG_BUF_LEN], path[PATH_MAX], *p; FILE *fp; // sscanf() is safe here, since send_ssi_file() also uses buffer // of size MG_BUF_LEN to get the tag. So strlen(tag) is always < MG_BUF_LEN. if (sscanf(tag, " virtual=\"%[^\"]\"", file_name) == 1) { // File name is relative to the webserver root (void) mg_snprintf(path, sizeof(path), "%s%c%s", conn->ctx->config[DOCUMENT_ROOT], '/', file_name); } else if (sscanf(tag, " abspath=\"%[^\"]\"", file_name) == 1) { // File name is relative to the webserver working directory // or it is absolute system path (void) mg_snprintf(path, sizeof(path), "%s", file_name); } else if (sscanf(tag, " file=\"%[^\"]\"", file_name) == 1 || sscanf(tag, " \"%[^\"]\"", file_name) == 1) { // File name is relative to the currect document (void) mg_snprintf(path, sizeof(path), "%s", ssi); if ((p = strrchr(path, '/')) != NULL) { p[1] = '\0'; } (void) mg_snprintf(path + strlen(path), sizeof(path) - strlen(path), "%s", file_name); } else { cry(conn, "Bad SSI #include: [%s]", tag); return; } if ((fp = mg_fopen(path, "rb")) == NULL) { cry(conn, "Cannot open SSI #include: [%s]: fopen(%s): %s", tag, path, strerror(ERRNO)); } else { fclose_on_exec(fp); if (match_prefix(conn->ctx->config[SSI_EXTENSIONS], strlen(conn->ctx->config[SSI_EXTENSIONS]), path) > 0) { send_ssi_file(conn, path, fp, include_level + 1); } else { send_file_data(conn, fp, 0, INT64_MAX); } fclose(fp); } } #if !defined(NO_POPEN) static void do_ssi_exec(struct mg_connection *conn, char *tag) { char cmd[MG_BUF_LEN]; FILE *fp; if (sscanf(tag, " \"%[^\"]\"", cmd) != 1) { cry(conn, "Bad SSI #exec: [%s]", tag); } else if ((fp = popen(cmd, "r")) == NULL) { cry(conn, "Cannot SSI #exec: [%s]: %s", cmd, strerror(ERRNO)); } else { send_file_data(conn, fp, 0, INT64_MAX); pclose(fp); } } #endif // !NO_POPEN static void send_ssi_file(struct mg_connection *conn, const char *path, FILE *fp, int include_level) { char buf[MG_BUF_LEN]; int ch, offset, len, in_ssi_tag; if (include_level > 10) { cry(conn, "SSI #include level is too deep (%s)", path); return; } in_ssi_tag = len = offset = 0; while ((ch = fgetc(fp)) != EOF) { if (in_ssi_tag && ch == '>') { in_ssi_tag = 0; buf[len++] = (char) ch; buf[len] = '\0'; assert(len <= (int) sizeof(buf)); if (len < 6 || memcmp(buf, "<!--#", 5) != 0) { // Not an SSI tag, pass it (void) mg_write(conn, buf, (size_t) len); } else { if (!memcmp(buf + 5, "include", 7)) { do_ssi_include(conn, path, buf + 12, include_level); #if !defined(NO_POPEN) } else if (!memcmp(buf + 5, "exec", 4)) { do_ssi_exec(conn, buf + 9); #endif // !NO_POPEN } else { cry(conn, "%s: unknown SSI " "command: \"%s\"", path, buf); } } len = 0; } else if (in_ssi_tag) { if (len == 5 && memcmp(buf, "<!--#", 5) != 0) { // Not an SSI tag in_ssi_tag = 0; } else if (len == (int) sizeof(buf) - 2) { cry(conn, "%s: SSI tag is too large", path); len = 0; } buf[len++] = ch & 0xff; } else if (ch == '<') { in_ssi_tag = 1; if (len > 0) { mg_write(conn, buf, (size_t) len); } len = 0; buf[len++] = ch & 0xff; } else { buf[len++] = ch & 0xff; if (len == (int) sizeof(buf)) { mg_write(conn, buf, (size_t) len); len = 0; } } } // Send the rest of buffered data if (len > 0) { mg_write(conn, buf, (size_t) len); } } static void handle_ssi_file_request(struct mg_connection *conn, const char *path) { struct vec mime_vec; FILE *fp; if ((fp = mg_fopen(path, "rb")) == NULL) { send_http_error(conn, 500, http_500_error, "fopen(%s): %s", path, strerror(ERRNO)); } else { conn->must_close = 1; fclose_on_exec(fp); get_mime_type(conn->ctx, path, &mime_vec); mg_printf(conn, "HTTP/1.1 200 OK\r\n" "Content-Type: %.*s\r\n" "Connection: close\r\n\r\n", (int) mime_vec.len, mime_vec.ptr); send_ssi_file(conn, path, fp, 0); fclose(fp); } } static void handle_options_request(struct mg_connection *conn) { static const char reply[] = "HTTP/1.1 200 OK\r\n" "Allow: GET, POST, HEAD, CONNECT, PUT, DELETE, OPTIONS, PROPFIND, MKCOL\r\n" "DAV: 1\r\n\r\n"; conn->status_code = 200; mg_write(conn, reply, sizeof(reply) - 1); } static int isbyte(int n) { return n >= 0 && n <= 255; } static int parse_net(const char *spec, uint32_t *net, uint32_t *mask) { int n, a, b, c, d, slash = 32, len = 0; if ((sscanf(spec, "%d.%d.%d.%d/%d%n", &a, &b, &c, &d, &slash, &n) == 5 || sscanf(spec, "%d.%d.%d.%d%n", &a, &b, &c, &d, &n) == 4) && isbyte(a) && isbyte(b) && isbyte(c) && isbyte(d) && slash >= 0 && slash < 33) { len = n; *net = ((uint32_t)a << 24) | ((uint32_t)b << 16) | ((uint32_t)c << 8) | d; *mask = slash ? 0xffffffffU << (32 - slash) : 0; } return len; } static int is_put_or_delete_request(const struct mg_connection *conn) { const char *s = conn->request_info.request_method; return s != NULL && (!strcmp(s, "PUT") || !strcmp(s, "DELETE") || !strcmp(s, "MKCOL")); } static int get_first_ssl_listener_index(const struct mg_context *ctx) { int i, index = -1; for (i = 0; index == -1 && i < ctx->num_listening_sockets; i++) { index = ctx->listening_sockets[i].is_ssl ? i : -1; } return index; } static void redirect_to_https_port(struct mg_connection *conn, int ssl_index) { char host[1025]; const char *host_header; if ((host_header = mg_get_header(conn, "Host")) == NULL || sscanf(host_header, "%1024[^:]", host) == 0) { // Cannot get host from the Host: header. Fallback to our IP address. sockaddr_to_string(host, sizeof(host), &conn->client.lsa); } mg_printf(conn, "HTTP/1.1 302 Found\r\nLocation: https://%s:%d%s\r\n\r\n", host, (int) ntohs(conn->ctx->listening_sockets[ssl_index]. lsa.sin.sin_port), conn->request_info.uri); } static void handle_delete_request(struct mg_connection *conn, const char *path) { struct file file = STRUCT_FILE_INITIALIZER; if (!mg_stat(path, &file)) { send_http_error(conn, 404, "Not Found", "%s", "File not found"); } else if (!file.modification_time) { send_http_error(conn, 500, http_500_error, "remove(%s): %s", path, strerror(ERRNO)); } else if (file.is_directory) { remove_directory(conn, path); send_http_error(conn, 204, "No Content", "%s", ""); } else if (mg_remove(path) == 0) { send_http_error(conn, 204, "No Content", "%s", ""); } else { send_http_error(conn, 423, "Locked", "remove(%s): %s", path, strerror(ERRNO)); } } // This is the heart of the Mongoose's logic. // This function is called when the request is read, parsed and validated, // and Mongoose must decide what action to take: serve a file, or // a directory, or call embedded function, etcetera. static void handle_request(struct mg_connection *conn) { struct mg_request_info *ri = &conn->request_info; char path[PATH_MAX]; int uri_len, ssl_index; struct file file = STRUCT_FILE_INITIALIZER; if ((conn->request_info.query_string = strchr(ri->uri, '?')) != NULL) { * ((char *) conn->request_info.query_string++) = '\0'; } uri_len = (int) strlen(ri->uri); mg_url_decode(ri->uri, uri_len, (char *) ri->uri, uri_len + 1, 0); remove_double_dots_and_double_slashes((char *) ri->uri); path[0] = '\0'; convert_uri_to_file_name(conn, path, sizeof(path), &file); // Perform redirect and auth checks before calling begin_request() handler. // Otherwise, begin_request() would need to perform auth checks and redirects. if (!conn->client.is_ssl && conn->client.ssl_redir && (ssl_index = get_first_ssl_listener_index(conn->ctx)) > -1) { redirect_to_https_port(conn, ssl_index); } else if (!is_put_or_delete_request(conn) && !check_authorization(conn, path)) { send_authorization_request(conn); } else if (call_user(MG_REQUEST_BEGIN, conn, (void *) ri->uri) == 1) { // Do nothing, callback has served the request } else if (!strcmp(ri->request_method, "OPTIONS")) { handle_options_request(conn); } else if (conn->ctx->config[DOCUMENT_ROOT] == NULL) { send_http_error(conn, 404, "Not Found", "Not Found"); } else if (is_put_or_delete_request(conn) && (is_authorized_for_put(conn) != 1)) { send_authorization_request(conn); } else if (!strcmp(ri->request_method, "PUT")) { put_file(conn, path); } else if (!strcmp(ri->request_method, "MKCOL")) { mkcol(conn, path); } else if (!strcmp(ri->request_method, "DELETE")) { handle_delete_request(conn, path); } else if (file.modification_time == (time_t) 0 || must_hide_file(conn, path)) { send_http_error(conn, 404, "Not Found", "%s", "File not found"); } else if (file.is_directory && ri->uri[uri_len - 1] != '/') { mg_printf(conn, "HTTP/1.1 301 Moved Permanently\r\n" "Location: %s/\r\n\r\n", ri->uri); } else if (!strcmp(ri->request_method, "PROPFIND")) { handle_propfind(conn, path, &file); } else if (file.is_directory && !substitute_index_file(conn, path, sizeof(path), &file)) { if (!mg_strcasecmp(conn->ctx->config[ENABLE_DIRECTORY_LISTING], "yes")) { handle_directory_request(conn, path); } else { send_http_error(conn, 403, "Directory Listing Denied", "Directory listing denied"); } #ifdef USE_LUA } else if (match_prefix("**.lp$", 6, path) > 0) { handle_lsp_request(conn, path, &file, NULL); #endif #if !defined(NO_CGI) } else if (match_prefix(conn->ctx->config[CGI_EXTENSIONS], strlen(conn->ctx->config[CGI_EXTENSIONS]), path) > 0) { if (strcmp(ri->request_method, "POST") && strcmp(ri->request_method, "HEAD") && strcmp(ri->request_method, "GET")) { send_http_error(conn, 501, "Not Implemented", "Method %s is not implemented", ri->request_method); } else { handle_cgi_request(conn, path); } #endif // !NO_CGI } else if (match_prefix(conn->ctx->config[SSI_EXTENSIONS], strlen(conn->ctx->config[SSI_EXTENSIONS]), path) > 0) { handle_ssi_file_request(conn, path); } else if (is_not_modified(conn, &file)) { send_http_error(conn, 304, "Not Modified", "%s", ""); } else { handle_file_request(conn, path, &file); } } static void close_all_listening_sockets(struct mg_context *ctx) { int i; for (i = 0; i < ctx->num_listening_sockets; i++) { closesocket(ctx->listening_sockets[i].sock); } free(ctx->listening_sockets); } static int is_valid_port(unsigned int port) { return port > 0 && port < 0xffff; } // Valid listening port specification is: [ip_address:]port[s] // Examples: 80, 443s, 127.0.0.1:3128, 1.2.3.4:8080s // TODO(lsm): add parsing of the IPv6 address static int parse_port_string(const struct vec *vec, struct socket *so) { unsigned int a, b, c, d, ch, port; int len; #if defined(USE_IPV6) char buf[100]; #endif // MacOS needs that. If we do not zero it, subsequent bind() will fail. // Also, all-zeroes in the socket address means binding to all addresses // for both IPv4 and IPv6 (INADDR_ANY and IN6ADDR_ANY_INIT). memset(so, 0, sizeof(*so)); so->lsa.sin.sin_family = AF_INET; if (sscanf(vec->ptr, "%u.%u.%u.%u:%u%n", &a, &b, &c, &d, &port, &len) == 5) { // Bind to a specific IPv4 address, e.g. 192.168.1.5:8080 so->lsa.sin.sin_addr.s_addr = htonl((a << 24) | (b << 16) | (c << 8) | d); so->lsa.sin.sin_port = htons((uint16_t) port); #if defined(USE_IPV6) } else if (sscanf(vec->ptr, "[%49[^]]]:%d%n", buf, &port, &len) == 2 && inet_pton(AF_INET6, buf, &so->lsa.sin6.sin6_addr)) { // IPv6 address, e.g. [3ffe:2a00:100:7031::1]:8080 so->lsa.sin6.sin6_family = AF_INET6; so->lsa.sin6.sin6_port = htons((uint16_t) port); #endif } else if (sscanf(vec->ptr, "%u%n", &port, &len) == 1) { // If only port is specified, bind to IPv4, INADDR_ANY so->lsa.sin.sin_port = htons((uint16_t) port); } else { port = len = 0; // Parsing failure. Make port invalid. } ch = vec->ptr[len]; // Next character after the port number so->is_ssl = ch == 's'; so->ssl_redir = ch == 'r'; // Make sure the port is valid and vector ends with 's', 'r' or ',' return is_valid_port(port) && (ch == '\0' || ch == 's' || ch == 'r' || ch == ','); } static int set_ports_option(struct mg_context *ctx) { const char *list = ctx->config[LISTENING_PORTS]; int on = 1, success = 1; #if defined(USE_IPV6) int off = 0; #endif struct vec vec; struct socket so, *ptr; while (success && (list = next_option(list, &vec, NULL)) != NULL) { if (!parse_port_string(&vec, &so)) { cry(fc(ctx), "%s: %.*s: invalid port spec. Expecting list of: %s", __func__, (int) vec.len, vec.ptr, "[IP_ADDRESS:]PORT[s|r]"); success = 0; } else if (so.is_ssl && ctx->ssl_ctx == NULL) { cry(fc(ctx), "Cannot add SSL socket, is -ssl_certificate option set?"); success = 0; } else if ((so.sock = socket(so.lsa.sa.sa_family, SOCK_STREAM, 6)) == INVALID_SOCKET || // On Windows, SO_REUSEADDR is recommended only for // broadcast UDP sockets setsockopt(so.sock, SOL_SOCKET, SO_REUSEADDR, (void *) &on, sizeof(on)) != 0 || #if defined(USE_IPV6) (so.lsa.sa.sa_family == AF_INET6 && setsockopt(so.sock, IPPROTO_IPV6, IPV6_V6ONLY, (void *) &off, sizeof(off)) != 0) || #endif bind(so.sock, &so.lsa.sa, so.lsa.sa.sa_family == AF_INET ? sizeof(so.lsa.sin) : sizeof(so.lsa)) != 0 || listen(so.sock, SOMAXCONN) != 0) { cry(fc(ctx), "%s: cannot bind to %.*s: %d (%s)", __func__, (int) vec.len, vec.ptr, ERRNO, strerror(errno)); closesocket(so.sock); success = 0; } else if ((ptr = (struct socket *) realloc(ctx->listening_sockets, (ctx->num_listening_sockets + 1) * sizeof(ctx->listening_sockets[0]))) == NULL) { closesocket(so.sock); success = 0; } else { set_close_on_exec(so.sock); ctx->listening_sockets = ptr; ctx->listening_sockets[ctx->num_listening_sockets] = so; ctx->num_listening_sockets++; } } if (!success) { close_all_listening_sockets(ctx); } return success; } // Verify given socket address against the ACL. // Return -1 if ACL is malformed, 0 if address is disallowed, 1 if allowed. static int check_acl(struct mg_context *ctx, uint32_t remote_ip) { int allowed, flag; uint32_t net, mask; struct vec vec; const char *list = ctx->config[ACCESS_CONTROL_LIST]; // If any ACL is set, deny by default allowed = list == NULL ? '+' : '-'; while ((list = next_option(list, &vec, NULL)) != NULL) { flag = vec.ptr[0]; if ((flag != '+' && flag != '-') || parse_net(&vec.ptr[1], &net, &mask) == 0) { cry(fc(ctx), "%s: subnet must be [+|-]x.x.x.x[/x]", __func__); return -1; } if (net == (remote_ip & mask)) { allowed = flag; } } return allowed == '+'; } #if !defined(_WIN32) static int set_uid_option(struct mg_context *ctx) { struct passwd *pw; const char *uid = ctx->config[RUN_AS_USER]; int success = 0; if (uid == NULL) { success = 1; } else { if ((pw = getpwnam(uid)) == NULL) { cry(fc(ctx), "%s: unknown user [%s]", __func__, uid); } else if (setgid(pw->pw_gid) == -1) { cry(fc(ctx), "%s: setgid(%s): %s", __func__, uid, strerror(errno)); } else if (setuid(pw->pw_uid) == -1) { cry(fc(ctx), "%s: setuid(%s): %s", __func__, uid, strerror(errno)); } else { success = 1; } } return success; } #endif // !_WIN32 static int set_gpass_option(struct mg_context *ctx) { struct file file = STRUCT_FILE_INITIALIZER; const char *path = ctx->config[GLOBAL_PASSWORDS_FILE]; if (path != NULL && !mg_stat(path, &file)) { cry(fc(ctx), "Cannot open %s: %s", path, strerror(ERRNO)); return 0; } return 1; } static int set_acl_option(struct mg_context *ctx) { return check_acl(ctx, (uint32_t) 0x7f000001UL) != -1; } static void reset_per_request_attributes(struct mg_connection *conn) { conn->path_info = NULL; conn->num_bytes_sent = conn->num_bytes_read = 0; conn->status_code = -1; conn->must_close = conn->request_len = 0; } static void close_socket_gracefully(struct mg_connection *conn) { #if defined(_WIN32) char buf[MG_BUF_LEN]; int n; #endif struct linger linger; // Set linger option to avoid socket hanging out after close. This prevent // ephemeral port exhaust problem under high QPS. linger.l_onoff = 1; linger.l_linger = 1; setsockopt(conn->client.sock, SOL_SOCKET, SO_LINGER, (char *) &linger, sizeof(linger)); // Send FIN to the client shutdown(conn->client.sock, SHUT_WR); set_non_blocking_mode(conn->client.sock); #if defined(_WIN32) // Read and discard pending incoming data. If we do not do that and close the // socket, the data in the send buffer may be discarded. This // behaviour is seen on Windows, when client keeps sending data // when server decides to close the connection; then when client // does recv() it gets no data back. do { n = pull(NULL, conn, buf, sizeof(buf)); } while (n > 0); #endif // Now we know that our FIN is ACK-ed, safe to close closesocket(conn->client.sock); } static void close_connection(struct mg_connection *conn) { conn->must_close = 1; #ifndef NO_SSL if (conn->ssl != NULL) { // Run SSL_shutdown twice to ensure completly close SSL connection SSL_shutdown(conn->ssl); SSL_free(conn->ssl); conn->ssl = NULL; } #endif if (conn->client.sock != INVALID_SOCKET) { close_socket_gracefully(conn); conn->client.sock = INVALID_SOCKET; } } void mg_close_connection(struct mg_connection *conn) { #ifndef NO_SSL if (conn->client_ssl_ctx != NULL) { SSL_CTX_free((SSL_CTX *) conn->client_ssl_ctx); } #endif close_connection(conn); free(conn); } static int is_valid_uri(const char *uri) { // Conform to http://www.w3.org/Protocols/rfc2616/rfc2616-sec5.html#sec5.1.2 // URI can be an asterisk (*) or should start with slash. return uri[0] == '/' || (uri[0] == '*' && uri[1] == '\0'); } static const char *getreq(struct mg_connection *conn) { const char *cl, *ret = NULL; reset_per_request_attributes(conn); conn->request_len = read_request(NULL, conn, conn->buf, conn->buf_size, &conn->data_len); assert(conn->request_len < 0 || conn->data_len >= conn->request_len); if (conn->request_len == 0 && conn->data_len == conn->buf_size) { conn->status_code = 413; ret = "Request Entity Too Large"; } else if (conn->request_len <= 0) { conn->status_code = 0; ret = "Client closed connection"; } else if (parse_http_message(conn->buf, conn->buf_size, &conn->request_info) <= 0) { conn->status_code = 400; ret = "Bad Request"; } else { // Request is valid. Set content_len attribute by parsing Content-Length // If Content-Length is absent, set content_len to 0 if request is GET, // and set it to INT64_MAX otherwise. Setting to INT64_MAX instructs // mg_read() to read from the socket until socket is closed. // The reason for treating GET and POST/PUT differently is that libraries // like jquery do not set Content-Length in GET requests, and we don't // want mg_read() to hang waiting until socket is timed out. // See https://github.com/cesanta/mongoose/pull/121 for more. conn->content_len = INT64_MAX; if (!mg_strcasecmp(conn->request_info.request_method, "GET")) { conn->content_len = 0; } if ((cl = get_header(&conn->request_info, "Content-Length")) != NULL) { conn->content_len = strtoll(cl, NULL, 10); } conn->birth_time = time(NULL); } return ret; } static void process_new_connection(struct mg_connection *conn) { struct mg_request_info *ri = &conn->request_info; int keep_alive_enabled, keep_alive, discard_len; const char *msg = NULL; keep_alive_enabled = !strcmp(conn->ctx->config[ENABLE_KEEP_ALIVE], "yes"); keep_alive = 0; // Important: on new connection, reset the receiving buffer. Credit goes // to crule42. conn->data_len = 0; do { if ((msg = getreq(conn)) != NULL) { conn->must_close = 1; } else if (!is_valid_uri(conn->request_info.uri)) { msg = "Bad Request"; conn->status_code = 400; } else if (strcmp(ri->http_version, "1.0") && strcmp(ri->http_version, "1.1")) { msg = "Bad HTTP Version"; conn->status_code = 505; } if (msg != NULL) { // Do not send anything to the client on timeout // see https://github.com/cesanta/mongoose/issues/261 if (conn->status_code > 0) { send_http_error(conn, conn->status_code, msg, "%s", ""); } } else { handle_request(conn); call_user(MG_REQUEST_END, conn, (void *) (long) conn->status_code); log_access(conn); } if (ri->remote_user != NULL) { free((void *) ri->remote_user); // Important! When having connections with and without auth // would cause double free and then crash ri->remote_user = NULL; } // NOTE(lsm): order is important here. should_keep_alive() call // is using parsed request, which will be invalid after memmove's below. // Therefore, memorize should_keep_alive() result now for later use // in loop exit condition. keep_alive = conn->ctx->stop_flag == 0 && keep_alive_enabled && conn->content_len >= 0 && should_keep_alive(conn); // Discard all buffered data for this request discard_len = conn->content_len >= 0 && conn->request_len > 0 && conn->request_len + conn->content_len < (int64_t) conn->data_len ? (int) (conn->request_len + conn->content_len) : conn->data_len; assert(discard_len >= 0); memmove(conn->buf, conn->buf + discard_len, conn->data_len - discard_len); conn->data_len -= discard_len; assert(conn->data_len >= 0); assert(conn->data_len <= conn->buf_size); } while (keep_alive); } // Worker threads take accepted socket from the queue static int consume_socket(struct mg_context *ctx, struct socket *sp) { recv(ctx->ctl[1], (void *) sp, sizeof(*sp), 0); return !ctx->stop_flag; } static void *worker_thread(void *thread_func_param) { struct mg_context *ctx = (struct mg_context *) thread_func_param; struct mg_connection *conn; conn = (struct mg_connection *) calloc(1, sizeof(*conn) + MAX_REQUEST_SIZE); if (conn == NULL) { cry(fc(ctx), "%s", "Cannot create new connection struct, OOM"); } else { conn->buf_size = MAX_REQUEST_SIZE; conn->buf = (char *) (conn + 1); conn->ctx = ctx; conn->event.user_data = ctx->user_data; call_user(MG_THREAD_BEGIN, conn, NULL); // Call consume_socket() even when ctx->stop_flag > 0, to let it signal // sq_empty condvar to wake up the master waiting in produce_socket() while (consume_socket(ctx, &conn->client)) { conn->birth_time = time(NULL); // Fill in IP, port info early so even if SSL setup below fails, // error handler would have the corresponding info. // Thanks to Johannes Winkelmann for the patch. // TODO(lsm): Fix IPv6 case conn->request_info.remote_port = ntohs(conn->client.rsa.sin.sin_port); memcpy(&conn->request_info.remote_ip, &conn->client.rsa.sin.sin_addr.s_addr, 4); conn->request_info.remote_ip = ntohl(conn->request_info.remote_ip); conn->request_info.is_ssl = conn->client.is_ssl; if (!conn->client.is_ssl #ifndef NO_SSL || sslize(conn, conn->ctx->ssl_ctx, SSL_accept) #endif ) { process_new_connection(conn); } close_connection(conn); } call_user(MG_THREAD_END, conn, NULL); free(conn); } // Signal master that we're done with connection and exiting send(ctx->ctl[1], "x", 1, 0); DEBUG_TRACE(("exiting")); return NULL; } // Master thread adds accepted socket to a queue static void produce_socket(struct mg_context *ctx, const struct socket *sp) { send(ctx->ctl[0], (void *) sp, sizeof(*sp), 0); } static int set_sock_timeout(SOCKET sock, int milliseconds) { #ifdef _WIN32 DWORD t = milliseconds; #else struct timeval t; t.tv_sec = milliseconds / 1000; t.tv_usec = (milliseconds * 1000) % 1000000; #endif return setsockopt(sock, SOL_SOCKET, SO_RCVTIMEO, (void *) &t, sizeof(t)) || setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO, (void *) &t, sizeof(t)); } static void accept_new_connection(const struct socket *listener, struct mg_context *ctx) { struct socket so; char src_addr[IP_ADDR_STR_LEN]; socklen_t len = sizeof(so.rsa); int on = 1; if ((so.sock = accept(listener->sock, &so.rsa.sa, &len)) == INVALID_SOCKET) { } else if (!check_acl(ctx, ntohl(* (uint32_t *) &so.rsa.sin.sin_addr))) { sockaddr_to_string(src_addr, sizeof(src_addr), &so.rsa); cry(fc(ctx), "%s: %s is not allowed to connect", __func__, src_addr); closesocket(so.sock); } else { // Put so socket structure into the queue DEBUG_TRACE(("Accepted socket %d", (int) so.sock)); set_close_on_exec(so.sock); so.is_ssl = listener->is_ssl; so.ssl_redir = listener->ssl_redir; getsockname(so.sock, &so.lsa.sa, &len); // Set TCP keep-alive. This is needed because if HTTP-level keep-alive // is enabled, and client resets the connection, server won't get // TCP FIN or RST and will keep the connection open forever. With TCP // keep-alive, next keep-alive handshake will figure out that the client // is down and will close the server end. // Thanks to Igor Klopov who suggested the patch. setsockopt(so.sock, SOL_SOCKET, SO_KEEPALIVE, (void *) &on, sizeof(on)); set_sock_timeout(so.sock, atoi(ctx->config[REQUEST_TIMEOUT])); produce_socket(ctx, &so); } } static void *master_thread(void *thread_func_param) { struct mg_context *ctx = (struct mg_context *) thread_func_param; struct pollfd *pfd; int i; // Increase priority of the master thread #if defined(_WIN32) SetThreadPriority(GetCurrentThread(), THREAD_PRIORITY_ABOVE_NORMAL); #endif #if defined(ISSUE_317) struct sched_param sched_param; sched_param.sched_priority = sched_get_priority_max(SCHED_RR); pthread_setschedparam(pthread_self(), SCHED_RR, &sched_param); #endif call_user(MG_THREAD_BEGIN, fc(ctx), NULL); pfd = (struct pollfd *) calloc(ctx->num_listening_sockets, sizeof(pfd[0])); while (pfd != NULL && ctx->stop_flag == 0) { for (i = 0; i < ctx->num_listening_sockets; i++) { pfd[i].fd = ctx->listening_sockets[i].sock; pfd[i].events = POLLIN; } if (poll(pfd, ctx->num_listening_sockets, 200) > 0) { for (i = 0; i < ctx->num_listening_sockets; i++) { // NOTE(lsm): on QNX, poll() returns POLLRDNORM after the // successfull poll, and POLLIN is defined as (POLLRDNORM | POLLRDBAND) // Therefore, we're checking pfd[i].revents & POLLIN, not // pfd[i].revents == POLLIN. if (ctx->stop_flag == 0 && (pfd[i].revents & POLLIN)) { accept_new_connection(&ctx->listening_sockets[i], ctx); } } } } free(pfd); DEBUG_TRACE(("stopping workers")); // Stop signal received: somebody called mg_stop. Quit. close_all_listening_sockets(ctx); // Wakeup workers that are waiting for connections to handle. for (i = 0; i < ctx->num_threads; i++) { struct socket dummy; send(ctx->ctl[0], (void *) &dummy, sizeof(dummy), 0); } // Wait until all threads finish for (i = 0; i < ctx->num_threads; i++) { char ch; recv(ctx->ctl[0], &ch, 1, 0); } #if !defined(NO_SSL) uninitialize_ssl(ctx); #endif DEBUG_TRACE(("exiting")); call_user(MG_THREAD_END, fc(ctx), NULL); // Signal mg_stop() that we're done. // WARNING: This must be the very last thing this // thread does, as ctx becomes invalid after this line. ctx->stop_flag = 2; return NULL; } static void free_context(struct mg_context *ctx) { int i; // Deallocate config parameters for (i = 0; i < NUM_OPTIONS; i++) { if (ctx->config[i] != NULL) free(ctx->config[i]); } #ifndef NO_SSL // Deallocate SSL context if (ctx->ssl_ctx != NULL) { SSL_CTX_free(ctx->ssl_ctx); } if (ssl_mutexes != NULL) { free(ssl_mutexes); ssl_mutexes = NULL; } #endif // !NO_SSL // Deallocate context itself free(ctx); } void mg_stop(struct mg_context *ctx) { ctx->stop_flag = 1; // Wait until mg_fini() stops while (ctx->stop_flag != 2) { (void) mg_sleep(10); } free_context(ctx); #if defined(_WIN32) && !defined(__SYMBIAN32__) (void) WSACleanup(); #endif // _WIN32 } struct mg_context *mg_start(const char **options, mg_event_handler_t func, void *user_data) { struct mg_context *ctx; const char *name, *value, *default_value; int i; #if defined(_WIN32) && !defined(__SYMBIAN32__) WSADATA data; WSAStartup(MAKEWORD(2,2), &data); #endif // _WIN32 // Allocate context and initialize reasonable general case defaults. // TODO(lsm): do proper error handling here. if ((ctx = (struct mg_context *) calloc(1, sizeof(*ctx))) == NULL) { return NULL; } ctx->event_handler = func; ctx->user_data = user_data; while (options && (name = *options++) != NULL) { if ((i = get_option_index(name)) == -1) { cry(fc(ctx), "Invalid option: %s", name); free_context(ctx); return NULL; } else if ((value = *options++) == NULL) { cry(fc(ctx), "%s: option value cannot be NULL", name); free_context(ctx); return NULL; } if (ctx->config[i] != NULL) { cry(fc(ctx), "warning: %s: duplicate option", name); free(ctx->config[i]); } ctx->config[i] = mg_strdup(value); DEBUG_TRACE(("[%s] -> [%s]", name, value)); } // Set default value if needed for (i = 0; config_options[i * 2] != NULL; i++) { default_value = config_options[i * 2 + 1]; if (ctx->config[i] == NULL && default_value != NULL) { ctx->config[i] = mg_strdup(default_value); } } // NOTE(lsm): order is important here. SSL certificates must // be initialized before listening ports. UID must be set last. if (!set_gpass_option(ctx) || #if !defined(NO_SSL) !set_ssl_option(ctx) || #endif !set_ports_option(ctx) || #if !defined(_WIN32) !set_uid_option(ctx) || #endif !set_acl_option(ctx)) { free_context(ctx); return NULL; } #if !defined(_WIN32) && !defined(__SYMBIAN32__) // Ignore SIGPIPE signal, so if browser cancels the request, it // won't kill the whole process. (void) signal(SIGPIPE, SIG_IGN); #endif // !_WIN32 mg_socketpair(ctx->ctl); // Start master (listening) thread mg_start_thread(master_thread, ctx); // Start worker threads for (i = 0; i < atoi(ctx->config[NUM_THREADS]); i++) { if (mg_start_thread(worker_thread, ctx) != 0) { cry(fc(ctx), "Cannot start worker thread: %ld", (long) ERRNO); } else { ctx->num_threads++; } } return ctx; }